Designing a Fault-Tolerant Distributed Computing Framework for Seismic Data Processing in Oil and Gas Exploration in the Niger Delta

Seismic data processing for oil and gas exploration in the Niger Delta generates extremely large datasets requiring parallel distributed computing, yet existing processing pipelines at Nigerian exploration companies are susceptible to single-node failures that cause costly computation restarts. This study designed a fault-tolerant distributed computing framework for seismic data processing workloads, validated on seismic migration and velocity analysis workflows. An Apache Spark-based distributed processing architecture was designed with Kubernetes container orchestration for deployment on a hybrid on-premise plus cloud cluster. A fault-tolerance mechanism was implemented using checkpoint-and-restart with HDFS-based intermediate data persistence, enabling processing recovery from any single node failure within 4 minutes without full job restart. The framework was benchmarked using a 2 TB synthetic 3D seismic dataset modelled on Niger Delta acquisition geometry. Processing throughput achieved 180 GB/hour on a 10-node cluster, compared to 24 GB/hour on a single workstation baseline. Recovery from a simulated node failure during Kirchhoff pre-stack depth migration reduced effective computation loss to 6 minutes, compared to a 4.2-hour full restart without fault tolerance. Cloud burst capability using AWS EC2 spot instances reduced peak computation cost by 61 percent compared to sustained on-premise provisioning. The study recommends the Nigerian National Petroleum Corporation Limited evaluate the framework for its exploration technology directorate and partner with the University of Port Harcourt for seismic algorithm optimisation research.

Keywords: distributed computing, fault tolerance, seismic data processing, Niger Delta, Apache Spark